{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:5BFNLN7C4LC4HESE3CFA5YC2KY","short_pith_number":"pith:5BFNLN7C","schema_version":"1.0","canonical_sha256":"e84ad5b7e2e2c5c39244d88a0ee05a561f26962d1eb6b585ed30fe710cd7d0dc","source":{"kind":"arxiv","id":"2605.29495","version":1},"attestation_state":"computed","paper":{"title":"On-Policy Replay for Continual Supervised Fine-Tuning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dongyang Xu, Jiaqi Huang, Meng Zhang, Taojie Zhu, Xin Chen, Yan Chen, Yizhi Wang","submitted_at":"2026-05-28T07:19:47Z","abstract_excerpt":"Continual supervised fine-tuning (SFT) is the de facto recipe for adapting large language models (LLMs) to a stream of downstream tasks, but it suffers from catastrophic forgetting of earlier capabilities. Recent work shows that on-policy signals -- training on the model's own outputs -- reduce forgetting more reliably than off-policy supervision. Existing on-policy methods route this signal through a new training objective (e.g., self-distillation losses with a teacher copy), inheriting an extra forward pass, schedule sensitivity, and stylistic drift from the teacher.We instead route the on-p"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.29495","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-28T07:19:47Z","cross_cats_sorted":[],"title_canon_sha256":"0e134535d8327a7234acd8a84df314fb28a8a41dc8584a0f6f32d37d9f3d8a40","abstract_canon_sha256":"23977264749dc4bbcff2594012f796ddb2550b06b2f0f3449e304b395aa350fc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T01:05:42.750852Z","signature_b64":"oqkIAJXt7q1k62Z08tBfFU4L3HL9i2O9yHqTY7RoL6Z+SxDuahY021Yd8bWOuSNuZ9Fc2tDnpj0y1+Lh2W0vCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e84ad5b7e2e2c5c39244d88a0ee05a561f26962d1eb6b585ed30fe710cd7d0dc","last_reissued_at":"2026-05-29T01:05:42.750328Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T01:05:42.750328Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On-Policy Replay for Continual Supervised Fine-Tuning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dongyang Xu, Jiaqi Huang, Meng Zhang, Taojie Zhu, Xin Chen, Yan Chen, Yizhi Wang","submitted_at":"2026-05-28T07:19:47Z","abstract_excerpt":"Continual supervised fine-tuning (SFT) is the de facto recipe for adapting large language models (LLMs) to a stream of downstream tasks, but it suffers from catastrophic forgetting of earlier capabilities. Recent work shows that on-policy signals -- training on the model's own outputs -- reduce forgetting more reliably than off-policy supervision. Existing on-policy methods route this signal through a new training objective (e.g., self-distillation losses with a teacher copy), inheriting an extra forward pass, schedule sensitivity, and stylistic drift from the teacher.We instead route the on-p"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.29495","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.29495/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.29495","created_at":"2026-05-29T01:05:42.750402+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.29495v1","created_at":"2026-05-29T01:05:42.750402+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.29495","created_at":"2026-05-29T01:05:42.750402+00:00"},{"alias_kind":"pith_short_12","alias_value":"5BFNLN7C4LC4","created_at":"2026-05-29T01:05:42.750402+00:00"},{"alias_kind":"pith_short_16","alias_value":"5BFNLN7C4LC4HESE","created_at":"2026-05-29T01:05:42.750402+00:00"},{"alias_kind":"pith_short_8","alias_value":"5BFNLN7C","created_at":"2026-05-29T01:05:42.750402+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5BFNLN7C4LC4HESE3CFA5YC2KY","json":"https://pith.science/pith/5BFNLN7C4LC4HESE3CFA5YC2KY.json","graph_json":"https://pith.science/api/pith-number/5BFNLN7C4LC4HESE3CFA5YC2KY/graph.json","events_json":"https://pith.science/api/pith-number/5BFNLN7C4LC4HESE3CFA5YC2KY/events.json","paper":"https://pith.science/paper/5BFNLN7C"},"agent_actions":{"view_html":"https://pith.science/pith/5BFNLN7C4LC4HESE3CFA5YC2KY","download_json":"https://pith.science/pith/5BFNLN7C4LC4HESE3CFA5YC2KY.json","view_paper":"https://pith.science/paper/5BFNLN7C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.29495&json=true","fetch_graph":"https://pith.science/api/pith-number/5BFNLN7C4LC4HESE3CFA5YC2KY/graph.json","fetch_events":"https://pith.science/api/pith-number/5BFNLN7C4LC4HESE3CFA5YC2KY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5BFNLN7C4LC4HESE3CFA5YC2KY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5BFNLN7C4LC4HESE3CFA5YC2KY/action/storage_attestation","attest_author":"https://pith.science/pith/5BFNLN7C4LC4HESE3CFA5YC2KY/action/author_attestation","sign_citation":"https://pith.science/pith/5BFNLN7C4LC4HESE3CFA5YC2KY/action/citation_signature","submit_replication":"https://pith.science/pith/5BFNLN7C4LC4HESE3CFA5YC2KY/action/replication_record"}},"created_at":"2026-05-29T01:05:42.750402+00:00","updated_at":"2026-05-29T01:05:42.750402+00:00"}